Horton

Re: [patch] BUG #15005: ANALYZE can make pg_class.reltuples inaccurate.

daveg <daveg@sonic.net>

View thread View on postgr.es
From: David Gould <daveg@sonic.net>
To: Justin Pryzby <pryzby@telsasoft.com>
Cc: Tomas Vondra <tomas.vondra@2ndquadrant.com>, pgsql-hackers@postgresql.org
Date: 2018-04-24T03:17:23Z
Lists: pgsql-hackers 
On Mon, 23 Apr 2018 20:09:21 -0500
Justin Pryzby <pryzby@telsasoft.com> wrote:

> Just want to add for the archive that I happened to run across what appears to
> be a 7-year old report of (I think) both of these vacuum/analyze bugs:
> 
> Re: [patch] BUG #15005: ANALYZE can make pg_class.reltuples inaccurate.                                                                                                                                            
> Re: [HACKERS] VACUUM and ANALYZE disagreeing on what reltuples means                                                                                                                                           
> 
> https://www.postgresql.org/message-id/flat/6373488.29.1323717222303.JavaMail.root%40store1.zcs.ext.wpsrv.net#6373488.29.1323717222303.JavaMail.root@store1.zcs.ext.wpsrv.net

Nice find. It does look like both, although, since the info is not in the
thread it is not certain that relpages was large enough to hit the analyze
issue. Unless that was with the old default statistics target in which case
it would be pretty easy to hit.

-dg


-- 
David Gould                                   daveg@sonic.net
If simplicity worked, the world would be overrun with insects.

Commits

  1. Sync up our various ways of estimating pg_class.reltuples.

  2. Fix tuple counting in SP-GiST index build.

  3. Fix errors in contrib/bloom index build.

  4. When updating reltuples after ANALYZE, just extrapolate from our sample.

  5. Avoid holding AutovacuumScheduleLock while rechecking table statistics.